Doug Kerr
Well-known member
Imagine that we are shooting a scene illuminated by "standard illuminant A" (a defined illuminant with a certain spectrum) and have the camera deliver the image in the sRGB color space, which has a "white point" (sometimes called its "reference white") which is the chromaticity of a different standard illuminant, illuminant D65.
We have the camera observe the scene illumination (using a white balance measurement diffuser, for example) and ask the camera to use that as its premise to make a white balance color correction. This shifts the chromaticities of each pixel in the image from what was directly implied by the sensor outputs.
We call this "white balance color correction". The process is an application of the broader topic called (and there are several other things that carry this same name), chromatic adaptation.
We can look at this in (at least) three ways:
A. In the image, we have replaced the colors as seen by the sensor with the colors that the sensor would have seen if the scene had been illuminated by illuminant D65.
B. In the image, we have made the colors of scene items such that, presented in a display, they would have the same appearance to an eye adapted to a surrounding illumination of illuminant D65 that they would have had to an eye, adapted to a surrounding illumination of illuminant A, viewing the scene itself.
C. We have transformed the color from representation "with respect to a reference illumination of illuminant A" to representation "with respect to a reference illumination of illuminant D65".
My own best grasp of this is that, in the case of white balance color correction, we have actually done A, and our purpose is to achieve B.
But color wonks often prefer outlook C (still recognizing that often our concern with the matter is to achieve B). The argument is that this is the most generalized outlook for the broad operation; for example, the scenario on which outlook "A" is predicated does not actually occur in many other situations of interest.
Unfortunately, there are rarely any tutorial explanations to help us grasp "outlook C". Merely saying that we "transform the color from representation with respect to one reference illumination to representation with respect to another reference illumination" is expected to make the whole concept leap out, but it doesn't (at least to me).
I'm actually at the moment trying to get my own understanding of this to congeal, and when that happens, I'll present what will hopefully be a story that clarifies what this actually means.
We have the camera observe the scene illumination (using a white balance measurement diffuser, for example) and ask the camera to use that as its premise to make a white balance color correction. This shifts the chromaticities of each pixel in the image from what was directly implied by the sensor outputs.
We call this "white balance color correction". The process is an application of the broader topic called (and there are several other things that carry this same name), chromatic adaptation.
We can look at this in (at least) three ways:
A. In the image, we have replaced the colors as seen by the sensor with the colors that the sensor would have seen if the scene had been illuminated by illuminant D65.
B. In the image, we have made the colors of scene items such that, presented in a display, they would have the same appearance to an eye adapted to a surrounding illumination of illuminant D65 that they would have had to an eye, adapted to a surrounding illumination of illuminant A, viewing the scene itself.
C. We have transformed the color from representation "with respect to a reference illumination of illuminant A" to representation "with respect to a reference illumination of illuminant D65".
My own best grasp of this is that, in the case of white balance color correction, we have actually done A, and our purpose is to achieve B.
But color wonks often prefer outlook C (still recognizing that often our concern with the matter is to achieve B). The argument is that this is the most generalized outlook for the broad operation; for example, the scenario on which outlook "A" is predicated does not actually occur in many other situations of interest.
Unfortunately, there are rarely any tutorial explanations to help us grasp "outlook C". Merely saying that we "transform the color from representation with respect to one reference illumination to representation with respect to another reference illumination" is expected to make the whole concept leap out, but it doesn't (at least to me).
I'm actually at the moment trying to get my own understanding of this to congeal, and when that happens, I'll present what will hopefully be a story that clarifies what this actually means.